Highlights
Block or report heyixuan2
Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.
You must be logged in to block users.
Contact GitHub support about this user’s behavior. Learn more about reporting abuse.
Report abuse
I build AI products inside large organizations, where the model is rarely the hard part.
What decides the outcome is the gap between a working demo and a system people depend on. Forward-deployed work means owning that gap — and it's the work I'm looking for.
Enterprise applications I built at Mercedes-Benz as the primary developer. The business context and requirements are shaped with the team. These systems run internally, so only the concept is shared here — not the code, screenshots, or data.
An enterprise AI transformation platform for any process, in any function. From Unstructured Documents to Organizational Change. Parses source documents into a verified BPMN map, designs the AI-native future state, and issues the implementation plan and workforce impact. Every score and FTE figure carries its source, method, and approver. A live pilot with its first users, now being engineered toward production. |
A full-lifecycle talent acquisition platform. Tracked and Supported at Every Stage. Consolidates job descriptions, resumes, interview materials, and transcripts into a single candidate record: competency models, candidate comparison, generated interview questions, AI interviews, and AI-assisted scoring. On the roadmap: an AI recruiter that sources candidates rather than waiting for applications. It supports the hiring decision; it does not make it. |
Built on nights and weekends, for no reason except that I wanted them to exist. I hope you enjoy them. Fork one, break it, tell me I got something wrong — or just say hi. And if you're building something open that looks like fun, count me in.
An agent skill that turns a sentence into a finished 3D print. From an Idea to a Physical Object. Finds or generates the model, checks printability, repairs the mesh, previews the result. You review and start the print in Bambu Studio; the agent only watches. Featured by Bambu Lab in their 2026 Skill Showcase. MIT licensed, and it runs inside Claude Code, Codex, Cursor, and 60-plus other agents. |
An LSTM–Transformer ensemble estimating whether an A-share stock closes up tomorrow. Finding Signal in Financial Time Series. Roughly 3 million sequences from 5,800 stocks, with market, sector, margin, and northbound-flow features. Ten models train on walk-forward splits, and permutation importance sorts the useful features from the harmful ones. A modeling and evaluation project, not a claim of investment performance. |
Pinned Loading
If the problem persists, check the GitHub status page or contact support.